Improving the Performance of Heart Disease Diagnosis by Combining the Cochleogram Transformation and Variable Auto-Encoder (VAE) Network

Ramin Barati ¹ ( Department of Electrical Engineering, Shiraz Branch, Islamic Azad University, Shiraz, Iran )
Mahbubeh Bahrayni ² ( Department of Electrical Engineering, Shiraz Branch, Islamic Azad University, Shiraz, Iran )
Abbas Kamali ³ ( Department of Electrical Engineering, Shiraz Branch, Islamic Azad University, Shiraz, Iran )

Submited date : 2023-09-12 Accepted date : 2023-12-12

Keywords: Variable Auto-Encoder (VAE) Network, : Cochleogram Transformation, Heart Disease Diagnosis, Support Vector Machine and K-Nearest Neighbor (KNN),

Abstract :

Early detection of abnormal heart sounds can largely prevent sudden death caused by heart diseases. A low-cost and non-invasive method for detecting abnormal heart sounds is the use of PCG signal. In this article, after segmenting the heart sound signal, the two-dimensional representation of the signal is obtained by cochleogram transformation, then with the help of deep learning and variable autoencoder network, 4 final features are extracted from each signal. Finally, support vector machine and k-nearest neighbor with k-fold validation are used to classify the heart sound signal into one of the predetermined categories of normal and abnormal sound class. In this research, the Physionet database, which has 3482 heart sounds from a standard collection, is used to train and evaluate the proposed method. The best results of the proposed method for classifying the two classes of heart sounds are 99.55, 98.75 and 99.70 in terms of accuracy, sensitivity and specificity, which is the higher ability of the proposed method compared to other methods. This abnormal sound detection system can be used very usefully in rural health centers and small hospitals to help doctors without expertise to diagnose heart problems.

Highlights:

Gammatone filter bank extracts frequency info via cochleagram transformation.

Latent space from autoencoders provides effective, compact features for classification.

Autoencoders excel in unsupervised learning by compressing features efficiently.

Robust to noise, autoencoders improve accuracy using compact latent features.

References:

[1] Cardiovascular diseases (CVDs): WHO Fact Sheet. WHO (11 June 2021). https://www.who.int/en/news-room/fact-sheets/detail/cardiovascular-diseases-(cvds).
[2] E. Etchells, C. Bell and K. Robb, “Does this patient have an abnormal systolic murmur?,” Jama, vol. 277, no 7, pp. 564–571, 1997.
[3] S. Das, S. Pal and M. Mitra, “Deep learning approach of murmur detection using Cochleagram,” Biomedical Signal Processing and Control, vol. 1, no. 77, p. 103747, 2022, doi: 10.1016/j.bspc.2022.103747.
[4] S.K. Randhawa and M. Singh, “Classification of heart sound signals using multi-modal features,” Procedia Computer Science, vol. 58, pp. 165-171, 2015, doi: 10.1016/j.procs.2015.08.045 .
[5] F. Ghaderi, H. R. Mohseni and S. Sanei, "Localizing Heart Sounds in Respiratory Signals Using Singular Spectrum Analysis," in IEEE Transactions on Biomedical Engineering, vol. 58, no. 12, pp. 3360-3367, Dec. 2011, doi: 10.1109/TBME.2011.2162728.
[6] Y. Zeinali and S.T.A. Niaki, “Heart sound classification using signal processing and machine learning algorithms,” Machine Learning with Applications, vol. 7, p. 100206. 2022, doi: 10.1016/j.mlwa.2021.100206.
[7] F. Safara and A.R.A. Ramaiah, “RenyiBS: Renyi entropy basis selection from wavelet packet decomposition tree for phonocardiogram classification,” The Journal of Supercomputing, vol. 77, pp. 3710-3726, 2021, doi: 10.1007/s11227-020-03413-9.
[8] B. Daoud, K. Nayad, B. Braham and B. Messaoud, “Heart murmurs detection and characterization using wavelet analysis with Renyi entropy,” Journal of Mechanics in Medicine and Biology, vol. 17, no. 6, 2017, doi: 10.1142/S0219519417500932.
[9] E. Kay and A. Agarwal, "DropConnected neural network trained with diverse features for classifying heart sounds," Computing in Cardiology Conference (CinC), Vancouver, BC, Canada, 2016, pp. 617-620.
[10] G. Eslamizadeh and R. Barati, “Heart murmur detection based on wavelet transformation and a synergy between artifcial neural network and modified neighbor annealing methods,” Artif. Intell. Med., vol. 78, pp. 23-40, 2017, doi: 10.1016/j.artmed.2017.05.005.
[11] H. Li, Y. Ren, G. Zhang, R. Wang, J. Cui and W. Zhang, "Detection and Classification of Abnormities of First Heart Sound Using Empirical Wavelet Transform," in IEEE Access, vol. 7, pp. 139643-139652, 2019, doi: 10.1109/ACCESS.2019.2943705.
[12] H. Chowdhury, K. N. Poudel and Y. Hu, "Time-Frequency Analysis, Denoising, Compression, Segmentation, and Classification of PCG Signals," in IEEE Access, vol. 8, pp. 160882-160890, 2020, doi: 10.1109/ACCESS.2020.3020806.
[13] J. Chen, Y. Wang and D. Wang, "A Feature Study for Classification-Based Speech Separation at Low Signal-to-Noise Ratios," in IEEE/ACM Transactions on Audio, Speech, and Language Processing, vol. 22, no. 12, pp. 1993-2002, Dec. 2014, doi: 10.1109/TASLP.2014.2359159.
[14] V. Maknickas and A. Maknickas, “Recognition of normal–abnormal phonocardiographic signals using deep convolutional neural networks and mel-frequency spectral coeffcients,” Physiol. Meas., vol. 38, no. 8, pp. 1671-1684, 2017, doi: 10.1088/1361-6579/aa7841.
[15] M. Hamidi, H. Ghassemian and M. Imani, “Classifcation of heart sound signal using curve ftting and fractal dimension,” Biomed. Signal Process. Control, vol. 39, pp. 351–359, 2018, doi: 10.1016/j.bspc.2017.08.002.
[16] M.S. Ahmad, J. Mir, M.O. Ullah, M.L.U.R. Shahid and M.A. Syed, “An effcient heart murmur recognition and cardiovascular disorders classifcation system,” Australas. Phys. Eng. Sci. Med., vol. 42, no. 3,pp. 733–743, 2019, doi: 10.1007/s13246-019-00778-x.
[17] G. Yaseen, Y. Son and S. Kwon, “Classifcation of heart sound signal using multiple features,” Appl. Sci., vol. 8, no. 12, 2018, doi: 10.3390/app8122344.
[18] P.T. Krishnan, P. Balasubramanian and S. Umapathy, “Automated heart sound classifcation system from unsegmented phonocardiogram (PCG) using deep neural network,” Phys. Eng. Sci. Med., vol. 43, no. 2, pp. 505–515,2020, doi: 10.1007/s13246-020-00851-w.
[19] S.L. Oh, V. Jahmunah, C.P. Ooi, R.-S. Tan, E.J. Ciaccio, T. Yamakawa, M. Tanabe, M. Kobayashi and U. Rajendra Acharya, “Classifcation of heart sound signals using a novel deep WaveNet model,” Comput. Methods Programs Biomed., vol. 196, p. 105604, 2020, doi: 10.1016/j.cmpb.2020.105604.
[20] P. Dhar, S. Dutta and V. Mukherjee, “Cross-wavelet assisted convolution neural network (AlexNet) approach for phonocardiogram signals classifcation,” Biomed. Signal Process. Control., vol. 63, p. 102142, 2021, doi: 10.1016/j.bspc.2020.102142 .
[21] . P. Dominguez-Morales, A. F. Jimenez-Fernandez, M. J. Dominguez-Morales and G. Jimenez-Moreno, "Deep Neural Networks for the Recognition and Classification of Heart Murmurs Using Neuromorphic Auditory Sensors," in IEEE Transactions on Biomedical Circuits and Systems, vol. 12, no. 1, pp. 24-34, Feb. 2018, doi: 10.1109/TBCAS.2017.2751545.
[22] S. Das, s. Pal and M. Mitra, “Acoustic feature based unsupervised approach of heart sound event detection,” Comput Biol Med., vol. 126, p. 103990, 2020, doi: 10.1016/j.compbiomed.2020.103990.
[23] C. Liu, D. Springer, Q. Li, B. Moody, R.A. Juan, F.J. Chorro, F. Castells, J.M. Roig, I. Silva, A.E.W. Johnson, Z. Syed, S.E. Schmidt, C.D. Papadaniil, L. Hadjileontiadis, H. Naseri, A. Moukadem, A. Dieterlen, C. Brandt, H. Tang, M. Samieinasab, M. R. Samieinasab, R. Sameni, R.G. Mark and G.D. Clifford, “An open access database for the evaluation of heart sound algorithms,” Physiol. Meas., vol. 37, no. 12, pp. 2181–2213, 2016, doi: 10.1088/0967-3334/37/12/2181.
[24] A.L. Goldberger, L.A.N. Amaral, L. Glass, J.M. Hausdorff, P.C. Ivanov, R.G. Mark, J. E. Mietus, G.B. Moody, C.-K. Peng and H.E. Stanley, “PhysioBank, PhysioToolkit, and PhysioNet: components of a new research resource for complex physiologic signals,” Circulation, vol. 101, no. 23, 2000, doi: 10.1161/01.cir.101.23.e215.
[25] C. Xu, S.J. Chao, G. Jiang, X. Chen, Q. He and P. Xie, “Two-level multidomain feature extraction on sparse representation for motor imagery classifcation,” Biomedical Signal Processing and Control, vol. 62, p. 102160, 2020, doi: 10.1016/j.bspc.2020.102160.
[26] H. Chang and J. Yang, “Genetic-based feature selection for efcient motion imaging of a brain–computer interface framework,” Journal of neural engineering, vol. 15, no. 5, 2018, doi: 10.1088/1741-2552/aad567.
[27] P. Tan, X. Wang and Y. Wang, “Dimensionality reduction in evolutionary algorithms-based feature selection for motor imagery braincomputer interface,” Swarm and Evolutionary Computation, vol. 52, p. 100597, 2020, doi: 10.1016/j.swevo.2019.100597.
[28] Y. Baysal, S. Ketenci, I.H. Altas and T. Kayikcioglu, “Multi-objective symbiotic organism search algorithm for optimal feature selection in brain computer interfaces,” Expert Systems with Applications, vol. 165, no. 113907, 2021, doi: 10.1016/j.eswa.2020.113907.
[29] M. A. Nadoomi and S. Majid, “Using Ant Colony Algorithm and Pairwise Learning to Classify Attack in Intrusion Detection Systems,” Journal of Communication Engineering, vol. 9, no. 36, pp. 39–53, 2020 [in persian].
[30] I. Jamali, S. Javad Mirabedini and A. Haronabadi, “Offering a Model for Persian Texts Classify by Combination of Classification Methods,” Journal of Communication Engineering, vol. 10, no. 38, pp. 61-72, 2020 [in persian].
[31] F. Hajiani, Naser Parhizgar and Ahmad Keshavarz, “Hyperspectral Image Classification Using Low Rank Representation and Spectral-Spatial Information,” Journal of Southern Communication Engineering, vol. 11, no. 43, pp. 27-38, 2022, doi: 10.30495/jce.2022.689206 [in persian].
[32] M. Jalali and T. Sedghi, “Extraction of Multiple Hybrid Features to Reduce the Semantic Vacuum with the Semi-Supervised Classification,” ,” Journal of Southern Communication Engineering, vol. 12, no. 45, pp. 31-44, 2022, doi: 10.30495/jce.2022.691134 [in persian].